net/ipv6/ipv6_sockglue.c

   1 /*
   2  *      IPv6 BSD socket options interface
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Based on linux/net/ipv4/ip_sockglue.c
   9  *
  10  *      This program is free software; you can redistribute it and/or
  11  *      modify it under the terms of the GNU General Public License
  12  *      as published by the Free Software Foundation; either version
  13  *      2 of the License, or (at your option) any later version.
  14  *
  15  *      FIXME: Make the setsockopt code POSIX compliant: That is
  16  *
  17  *      o       Truncate getsockopt returns
  18  *      o       Return an optlen of the truncated length if need be
  19  *
  20  *      Changes:
  21  *      David L Stevens <dlstevens@us.ibm.com>:
  22  *              - added multicast source filtering API for MLDv2
  23  */
  24
  25 #include <linux/module.h>
  26 #include <linux/capability.h>
  27 #include <linux/errno.h>
  28 #include <linux/types.h>
  29 #include <linux/socket.h>
  30 #include <linux/sockios.h>
  31 #include <linux/net.h>
  32 #include <linux/in6.h>
  33 #include <linux/mroute6.h>
  34 #include <linux/netdevice.h>
  35 #include <linux/if_arp.h>
  36 #include <linux/init.h>
  37 #include <linux/sysctl.h>
  38 #include <linux/netfilter.h>
  39 #include <linux/slab.h>
  40
  41 #include <net/sock.h>
  42 #include <net/snmp.h>
  43 #include <net/ipv6.h>
  44 #include <net/ndisc.h>
  45 #include <net/protocol.h>
  46 #include <net/transp_v6.h>
  47 #include <net/ip6_route.h>
  48 #include <net/addrconf.h>
  49 #include <net/inet_common.h>
  50 #include <net/tcp.h>
  51 #include <net/udp.h>
  52 #include <net/udplite.h>
  53 #include <net/xfrm.h>
  54 #include <net/compat.h>
  55
  56 #include <asm/uaccess.h>
  57
  58 struct ip6_ra_chain *ip6_ra_chain;
  59 DEFINE_RWLOCK(ip6_ra_lock);
  60
  61 int ip6_ra_control(struct sock *sk, int sel)
  62 {
  63         struct ip6_ra_chain *ra, *new_ra, **rap;
  64
  65         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
  66         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
  67                 return -ENOPROTOOPT;
  68
  69         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
  70
  71         write_lock_bh(&ip6_ra_lock);
  72         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
  73                 if (ra->sk == sk) {
  74                         if (sel>=0) {
  75                                 write_unlock_bh(&ip6_ra_lock);
  76                                 kfree(new_ra);
  77                                 return -EADDRINUSE;
  78                         }
  79
  80                         *rap = ra->next;
  81                         write_unlock_bh(&ip6_ra_lock);
  82
  83                         sock_put(sk);
  84                         kfree(ra);
  85                         return 0;
  86                 }
  87         }
  88         if (new_ra == NULL) {
  89                 write_unlock_bh(&ip6_ra_lock);
  90                 return -ENOBUFS;
  91         }
  92         new_ra->sk = sk;
  93         new_ra->sel = sel;
  94         new_ra->next = ra;
  95         *rap = new_ra;
  96         sock_hold(sk);
  97         write_unlock_bh(&ip6_ra_lock);
  98         return 0;
  99 }
 100
 101 static
 102 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
 103                                            struct ipv6_txoptions *opt)
 104 {
 105         if (inet_sk(sk)->is_icsk) {
 106                 if (opt &&
 107                     !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
 108                     inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) {
 109                         struct inet_connection_sock *icsk = inet_csk(sk);
 110                         icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
 111                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 112                 }
 113                 opt = xchg(&inet6_sk(sk)->opt, opt);
 114         } else {
 115                 spin_lock(&sk->sk_dst_lock);
 116                 opt = xchg(&inet6_sk(sk)->opt, opt);
 117                 spin_unlock(&sk->sk_dst_lock);
 118         }
 119         sk_dst_reset(sk);
 120
 121         return opt;
 122 }
 123
 124 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 125                     char __user *optval, unsigned int optlen)
 126 {
 127         struct ipv6_pinfo *np = inet6_sk(sk);
 128         struct net *net = sock_net(sk);
 129         int val, valbool;
 130         int retv = -ENOPROTOOPT;
 131
 132         if (optval == NULL)
 133                 val=0;
 134         else {
 135                 if (optlen >= sizeof(int)) {
 136                         if (get_user(val, (int __user *) optval))
 137                                 return -EFAULT;
 138                 } else
 139                         val = 0;
 140         }
 141
 142         valbool = (val!=0);
 143
 144         if (ip6_mroute_opt(optname))
 145                 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
 146
 147         lock_sock(sk);
 148
 149         switch (optname) {
 150
 151         case IPV6_ADDRFORM:
 152                 if (optlen < sizeof(int))
 153                         goto e_inval;
 154                 if (val == PF_INET) {
 155                         struct ipv6_txoptions *opt;
 156                         struct sk_buff *pktopt;
 157
 158                         if (sk->sk_type == SOCK_RAW)
 159                                 break;
 160
 161                         if (sk->sk_protocol == IPPROTO_UDP ||
 162                             sk->sk_protocol == IPPROTO_UDPLITE) {
 163                                 struct udp_sock *up = udp_sk(sk);
 164                                 if (up->pending == AF_INET6) {
 165                                         retv = -EBUSY;
 166                                         break;
 167                                 }
 168                         } else if (sk->sk_protocol != IPPROTO_TCP)
 169                                 break;
 170
 171                         if (sk->sk_state != TCP_ESTABLISHED) {
 172                                 retv = -ENOTCONN;
 173                                 break;
 174                         }
 175
 176                         if (ipv6_only_sock(sk) ||
 177                             !ipv6_addr_v4mapped(&np->daddr)) {
 178                                 retv = -EADDRNOTAVAIL;
 179                                 break;
 180                         }
 181
 182                         fl6_free_socklist(sk);
 183                         ipv6_sock_mc_close(sk);
 184
 185                         /*
 186                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
 187                          * remove it from the refcnt debug socks count in the
 188                          * original family...
 189                          */
 190                         sk_refcnt_debug_dec(sk);
 191
 192                         if (sk->sk_protocol == IPPROTO_TCP) {
 193                                 struct inet_connection_sock *icsk = inet_csk(sk);
 194                                 local_bh_disable();
 195                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 196                                 sock_prot_inuse_add(net, &tcp_prot, 1);
 197                                 local_bh_enable();
 198                                 sk->sk_prot = &tcp_prot;
 199                                 icsk->icsk_af_ops = &ipv4_specific;
 200                                 sk->sk_socket->ops = &inet_stream_ops;
 201                                 sk->sk_family = PF_INET;
 202                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 203                         } else {
 204                                 struct proto *prot = &udp_prot;
 205
 206                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
 207                                         prot = &udplite_prot;
 208                                 local_bh_disable();
 209                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 210                                 sock_prot_inuse_add(net, prot, 1);
 211                                 local_bh_enable();
 212                                 sk->sk_prot = prot;
 213                                 sk->sk_socket->ops = &inet_dgram_ops;
 214                                 sk->sk_family = PF_INET;
 215                         }
 216                         opt = xchg(&np->opt, NULL);
 217                         if (opt)
 218                                 sock_kfree_s(sk, opt, opt->tot_len);
 219                         pktopt = xchg(&np->pktoptions, NULL);
 220                         kfree_skb(pktopt);
 221
 222                         sk->sk_destruct = inet_sock_destruct;
 223                         /*
 224                          * ... and add it to the refcnt debug socks count
 225                          * in the new family. -acme
 226                          */
 227                         sk_refcnt_debug_inc(sk);
 228                         module_put(THIS_MODULE);
 229                         retv = 0;
 230                         break;
 231                 }
 232                 goto e_inval;
 233
 234         case IPV6_V6ONLY:
 235                 if (optlen < sizeof(int) ||
 236                     inet_sk(sk)->inet_num)
 237                         goto e_inval;
 238                 np->ipv6only = valbool;
 239                 retv = 0;
 240                 break;
 241
 242         case IPV6_RECVPKTINFO:
 243                 if (optlen < sizeof(int))
 244                         goto e_inval;
 245                 np->rxopt.bits.rxinfo = valbool;
 246                 retv = 0;
 247                 break;
 248
 249         case IPV6_2292PKTINFO:
 250                 if (optlen < sizeof(int))
 251                         goto e_inval;
 252                 np->rxopt.bits.rxoinfo = valbool;
 253                 retv = 0;
 254                 break;
 255
 256         case IPV6_RECVHOPLIMIT:
 257                 if (optlen < sizeof(int))
 258                         goto e_inval;
 259                 np->rxopt.bits.rxhlim = valbool;
 260                 retv = 0;
 261                 break;
 262
 263         case IPV6_2292HOPLIMIT:
 264                 if (optlen < sizeof(int))
 265                         goto e_inval;
 266                 np->rxopt.bits.rxohlim = valbool;
 267                 retv = 0;
 268                 break;
 269
 270         case IPV6_RECVRTHDR:
 271                 if (optlen < sizeof(int))
 272                         goto e_inval;
 273                 np->rxopt.bits.srcrt = valbool;
 274                 retv = 0;
 275                 break;
 276
 277         case IPV6_2292RTHDR:
 278                 if (optlen < sizeof(int))
 279                         goto e_inval;
 280                 np->rxopt.bits.osrcrt = valbool;
 281                 retv = 0;
 282                 break;
 283
 284         case IPV6_RECVHOPOPTS:
 285                 if (optlen < sizeof(int))
 286                         goto e_inval;
 287                 np->rxopt.bits.hopopts = valbool;
 288                 retv = 0;
 289                 break;
 290
 291         case IPV6_2292HOPOPTS:
 292                 if (optlen < sizeof(int))
 293                         goto e_inval;
 294                 np->rxopt.bits.ohopopts = valbool;
 295                 retv = 0;
 296                 break;
 297
 298         case IPV6_RECVDSTOPTS:
 299                 if (optlen < sizeof(int))
 300                         goto e_inval;
 301                 np->rxopt.bits.dstopts = valbool;
 302                 retv = 0;
 303                 break;
 304
 305         case IPV6_2292DSTOPTS:
 306                 if (optlen < sizeof(int))
 307                         goto e_inval;
 308                 np->rxopt.bits.odstopts = valbool;
 309                 retv = 0;
 310                 break;
 311
 312         case IPV6_TCLASS:
 313                 if (optlen < sizeof(int))
 314                         goto e_inval;
 315                 if (val < -1 || val > 0xff)
 316                         goto e_inval;
 317                 /* RFC 3542, 6.5: default traffic class of 0x0 */
 318                 if (val == -1)
 319                         val = 0;
 320                 np->tclass = val;
 321                 retv = 0;
 322                 break;
 323
 324         case IPV6_RECVTCLASS:
 325                 if (optlen < sizeof(int))
 326                         goto e_inval;
 327                 np->rxopt.bits.rxtclass = valbool;
 328                 retv = 0;
 329                 break;
 330
 331         case IPV6_FLOWINFO:
 332                 if (optlen < sizeof(int))
 333                         goto e_inval;
 334                 np->rxopt.bits.rxflow = valbool;
 335                 retv = 0;
 336                 break;
 337
 338         case IPV6_RECVPATHMTU:
 339                 if (optlen < sizeof(int))
 340                         goto e_inval;
 341                 np->rxopt.bits.rxpmtu = valbool;
 342                 retv = 0;
 343                 break;
 344
 345         case IPV6_TRANSPARENT:
 346                 if (optlen < sizeof(int))
 347                         goto e_inval;
 348                 /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
 349                 inet_sk(sk)->transparent = valbool;
 350                 retv = 0;
 351                 break;
 352
 353         case IPV6_RECVORIGDSTADDR:
 354                 if (optlen < sizeof(int))
 355                         goto e_inval;
 356                 np->rxopt.bits.rxorigdstaddr = valbool;
 357                 retv = 0;
 358                 break;
 359
 360         case IPV6_HOPOPTS:
 361         case IPV6_RTHDRDSTOPTS:
 362         case IPV6_RTHDR:
 363         case IPV6_DSTOPTS:
 364         {
 365                 struct ipv6_txoptions *opt;
 366
 367                 /* remove any sticky options header with a zero option
 368                  * length, per RFC3542.
 369                  */
 370                 if (optlen == 0)
 371                         optval = NULL;
 372                 else if (optval == NULL)
 373                         goto e_inval;
 374                 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
 375                          optlen & 0x7 || optlen > 8 * 255)
 376                         goto e_inval;
 377
 378                 /* hop-by-hop / destination options are privileged option */
 379                 retv = -EPERM;
 380                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
 381                         break;
 382
 383                 opt = ipv6_renew_options(sk, np->opt, optname,
 384                                          (struct ipv6_opt_hdr __user *)optval,
 385                                          optlen);
 386                 if (IS_ERR(opt)) {
 387                         retv = PTR_ERR(opt);
 388                         break;
 389                 }
 390
 391                 /* routing header option needs extra check */
 392                 retv = -EINVAL;
 393                 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
 394                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
 395                         switch (rthdr->type) {
 396 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 397                         case IPV6_SRCRT_TYPE_2:
 398                                 if (rthdr->hdrlen != 2 ||
 399                                     rthdr->segments_left != 1)
 400                                         goto sticky_done;
 401
 402                                 break;
 403 #endif
 404                         default:
 405                                 goto sticky_done;
 406                         }
 407                 }
 408
 409                 retv = 0;
 410                 opt = ipv6_update_options(sk, opt);
 411 sticky_done:
 412                 if (opt)
 413                         sock_kfree_s(sk, opt, opt->tot_len);
 414                 break;
 415         }
 416
 417         case IPV6_PKTINFO:
 418         {
 419                 struct in6_pktinfo pkt;
 420
 421                 if (optlen == 0)
 422                         goto e_inval;
 423                 else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
 424                         goto e_inval;
 425
 426                 if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
 427                                 retv = -EFAULT;
 428                                 break;
 429                 }
 430                 if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
 431                         goto e_inval;
 432
 433                 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
 434                 ipv6_addr_copy(&np->sticky_pktinfo.ipi6_addr, &pkt.ipi6_addr);
 435                 retv = 0;
 436                 break;
 437         }
 438
 439         case IPV6_2292PKTOPTIONS:
 440         {
 441                 struct ipv6_txoptions *opt = NULL;
 442                 struct msghdr msg;
 443                 struct flowi fl;
 444                 int junk;
 445
 446                 fl.fl6_flowlabel = 0;
 447                 fl.oif = sk->sk_bound_dev_if;
 448                 fl.mark = sk->sk_mark;
 449
 450                 if (optlen == 0)
 451                         goto update;
 452
 453                 /* 1K is probably excessive
 454                  * 1K is surely not enough, 2K per standard header is 16K.
 455                  */
 456                 retv = -EINVAL;
 457                 if (optlen > 64*1024)
 458                         break;
 459
 460                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 461                 retv = -ENOBUFS;
 462                 if (opt == NULL)
 463                         break;
 464
 465                 memset(opt, 0, sizeof(*opt));
 466                 opt->tot_len = sizeof(*opt) + optlen;
 467                 retv = -EFAULT;
 468                 if (copy_from_user(opt+1, optval, optlen))
 469                         goto done;
 470
 471                 msg.msg_controllen = optlen;
 472                 msg.msg_control = (void*)(opt+1);
 473
 474                 retv = datagram_send_ctl(net, &msg, &fl, opt, &junk, &junk,
 475                                          &junk);
 476                 if (retv)
 477                         goto done;
 478 update:
 479                 retv = 0;
 480                 opt = ipv6_update_options(sk, opt);
 481 done:
 482                 if (opt)
 483                         sock_kfree_s(sk, opt, opt->tot_len);
 484                 break;
 485         }
 486         case IPV6_UNICAST_HOPS:
 487                 if (optlen < sizeof(int))
 488                         goto e_inval;
 489                 if (val > 255 || val < -1)
 490                         goto e_inval;
 491                 np->hop_limit = val;
 492                 retv = 0;
 493                 break;
 494
 495         case IPV6_MULTICAST_HOPS:
 496                 if (sk->sk_type == SOCK_STREAM)
 497                         break;
 498                 if (optlen < sizeof(int))
 499                         goto e_inval;
 500                 if (val > 255 || val < -1)
 501                         goto e_inval;
 502                 np->mcast_hops = val;
 503                 retv = 0;
 504                 break;
 505
 506         case IPV6_MULTICAST_LOOP:
 507                 if (optlen < sizeof(int))
 508                         goto e_inval;
 509                 if (val != valbool)
 510                         goto e_inval;
 511                 np->mc_loop = valbool;
 512                 retv = 0;
 513                 break;
 514
 515         case IPV6_MULTICAST_IF:
 516                 if (sk->sk_type == SOCK_STREAM)
 517                         break;
 518                 if (optlen < sizeof(int))
 519                         goto e_inval;
 520
 521                 if (val) {
 522                         struct net_device *dev;
 523
 524                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
 525                                 goto e_inval;
 526
 527                         dev = dev_get_by_index(net, val);
 528                         if (!dev) {
 529                                 retv = -ENODEV;
 530                                 break;
 531                         }
 532                         dev_put(dev);
 533                 }
 534                 np->mcast_oif = val;
 535                 retv = 0;
 536                 break;
 537         case IPV6_ADD_MEMBERSHIP:
 538         case IPV6_DROP_MEMBERSHIP:
 539         {
 540                 struct ipv6_mreq mreq;
 541
 542                 if (optlen < sizeof(struct ipv6_mreq))
 543                         goto e_inval;
 544
 545                 retv = -EPROTO;
 546                 if (inet_sk(sk)->is_icsk)
 547                         break;
 548
 549                 retv = -EFAULT;
 550                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 551                         break;
 552
 553                 if (optname == IPV6_ADD_MEMBERSHIP)
 554                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 555                 else
 556                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 557                 break;
 558         }
 559         case IPV6_JOIN_ANYCAST:
 560         case IPV6_LEAVE_ANYCAST:
 561         {
 562                 struct ipv6_mreq mreq;
 563
 564                 if (optlen < sizeof(struct ipv6_mreq))
 565                         goto e_inval;
 566
 567                 retv = -EFAULT;
 568                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 569                         break;
 570
 571                 if (optname == IPV6_JOIN_ANYCAST)
 572                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 573                 else
 574                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 575                 break;
 576         }
 577         case MCAST_JOIN_GROUP:
 578         case MCAST_LEAVE_GROUP:
 579         {
 580                 struct group_req greq;
 581                 struct sockaddr_in6 *psin6;
 582
 583                 if (optlen < sizeof(struct group_req))
 584                         goto e_inval;
 585
 586                 retv = -EFAULT;
 587                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
 588                         break;
 589                 if (greq.gr_group.ss_family != AF_INET6) {
 590                         retv = -EADDRNOTAVAIL;
 591                         break;
 592                 }
 593                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 594                 if (optname == MCAST_JOIN_GROUP)
 595                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
 596                                 &psin6->sin6_addr);
 597                 else
 598                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
 599                                 &psin6->sin6_addr);
 600                 break;
 601         }
 602         case MCAST_JOIN_SOURCE_GROUP:
 603         case MCAST_LEAVE_SOURCE_GROUP:
 604         case MCAST_BLOCK_SOURCE:
 605         case MCAST_UNBLOCK_SOURCE:
 606         {
 607                 struct group_source_req greqs;
 608                 int omode, add;
 609
 610                 if (optlen < sizeof(struct group_source_req))
 611                         goto e_inval;
 612                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 613                         retv = -EFAULT;
 614                         break;
 615                 }
 616                 if (greqs.gsr_group.ss_family != AF_INET6 ||
 617                     greqs.gsr_source.ss_family != AF_INET6) {
 618                         retv = -EADDRNOTAVAIL;
 619                         break;
 620                 }
 621                 if (optname == MCAST_BLOCK_SOURCE) {
 622                         omode = MCAST_EXCLUDE;
 623                         add = 1;
 624                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 625                         omode = MCAST_EXCLUDE;
 626                         add = 0;
 627                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 628                         struct sockaddr_in6 *psin6;
 629
 630                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 631                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
 632                                 &psin6->sin6_addr);
 633                         /* prior join w/ different source is ok */
 634                         if (retv && retv != -EADDRINUSE)
 635                                 break;
 636                         omode = MCAST_INCLUDE;
 637                         add = 1;
 638                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 639                         omode = MCAST_INCLUDE;
 640                         add = 0;
 641                 }
 642                 retv = ip6_mc_source(add, omode, sk, &greqs);
 643                 break;
 644         }
 645         case MCAST_MSFILTER:
 646         {
 647                 extern int sysctl_mld_max_msf;
 648                 struct group_filter *gsf;
 649
 650                 if (optlen < GROUP_FILTER_SIZE(0))
 651                         goto e_inval;
 652                 if (optlen > sysctl_optmem_max) {
 653                         retv = -ENOBUFS;
 654                         break;
 655                 }
 656                 gsf = kmalloc(optlen,GFP_KERNEL);
 657                 if (!gsf) {
 658                         retv = -ENOBUFS;
 659                         break;
 660                 }
 661                 retv = -EFAULT;
 662                 if (copy_from_user(gsf, optval, optlen)) {
 663                         kfree(gsf);
 664                         break;
 665                 }
 666                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 667                 if (gsf->gf_numsrc >= 0x1ffffffU ||
 668                     gsf->gf_numsrc > sysctl_mld_max_msf) {
 669                         kfree(gsf);
 670                         retv = -ENOBUFS;
 671                         break;
 672                 }
 673                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 674                         kfree(gsf);
 675                         retv = -EINVAL;
 676                         break;
 677                 }
 678                 retv = ip6_mc_msfilter(sk, gsf);
 679                 kfree(gsf);
 680
 681                 break;
 682         }
 683         case IPV6_ROUTER_ALERT:
 684                 if (optlen < sizeof(int))
 685                         goto e_inval;
 686                 retv = ip6_ra_control(sk, val);
 687                 break;
 688         case IPV6_MTU_DISCOVER:
 689                 if (optlen < sizeof(int))
 690                         goto e_inval;
 691                 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
 692                         goto e_inval;
 693                 np->pmtudisc = val;
 694                 retv = 0;
 695                 break;
 696         case IPV6_MTU:
 697                 if (optlen < sizeof(int))
 698                         goto e_inval;
 699                 if (val && val < IPV6_MIN_MTU)
 700                         goto e_inval;
 701                 np->frag_size = val;
 702                 retv = 0;
 703                 break;
 704         case IPV6_RECVERR:
 705                 if (optlen < sizeof(int))
 706                         goto e_inval;
 707                 np->recverr = valbool;
 708                 if (!val)
 709                         skb_queue_purge(&sk->sk_error_queue);
 710                 retv = 0;
 711                 break;
 712         case IPV6_FLOWINFO_SEND:
 713                 if (optlen < sizeof(int))
 714                         goto e_inval;
 715                 np->sndflow = valbool;
 716                 retv = 0;
 717                 break;
 718         case IPV6_FLOWLABEL_MGR:
 719                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
 720                 break;
 721         case IPV6_IPSEC_POLICY:
 722         case IPV6_XFRM_POLICY:
 723                 retv = -EPERM;
 724                 if (!capable(CAP_NET_ADMIN))
 725                         break;
 726                 retv = xfrm_user_policy(sk, optname, optval, optlen);
 727                 break;
 728
 729         case IPV6_ADDR_PREFERENCES:
 730             {
 731                 unsigned int pref = 0;
 732                 unsigned int prefmask = ~0;
 733
 734                 if (optlen < sizeof(int))
 735                         goto e_inval;
 736
 737                 retv = -EINVAL;
 738
 739                 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
 740                 switch (val & (IPV6_PREFER_SRC_PUBLIC|
 741                                IPV6_PREFER_SRC_TMP|
 742                                IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
 743                 case IPV6_PREFER_SRC_PUBLIC:
 744                         pref |= IPV6_PREFER_SRC_PUBLIC;
 745                         break;
 746                 case IPV6_PREFER_SRC_TMP:
 747                         pref |= IPV6_PREFER_SRC_TMP;
 748                         break;
 749                 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
 750                         break;
 751                 case 0:
 752                         goto pref_skip_pubtmp;
 753                 default:
 754                         goto e_inval;
 755                 }
 756
 757                 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
 758                               IPV6_PREFER_SRC_TMP);
 759 pref_skip_pubtmp:
 760
 761                 /* check HOME/COA conflicts */
 762                 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
 763                 case IPV6_PREFER_SRC_HOME:
 764                         break;
 765                 case IPV6_PREFER_SRC_COA:
 766                         pref |= IPV6_PREFER_SRC_COA;
 767                 case 0:
 768                         goto pref_skip_coa;
 769                 default:
 770                         goto e_inval;
 771                 }
 772
 773                 prefmask &= ~IPV6_PREFER_SRC_COA;
 774 pref_skip_coa:
 775
 776                 /* check CGA/NONCGA conflicts */
 777                 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
 778                 case IPV6_PREFER_SRC_CGA:
 779                 case IPV6_PREFER_SRC_NONCGA:
 780                 case 0:
 781                         break;
 782                 default:
 783                         goto e_inval;
 784                 }
 785
 786                 np->srcprefs = (np->srcprefs & prefmask) | pref;
 787                 retv = 0;
 788
 789                 break;
 790             }
 791         case IPV6_MINHOPCOUNT:
 792                 if (optlen < sizeof(int))
 793                         goto e_inval;
 794                 if (val < 0 || val > 255)
 795                         goto e_inval;
 796                 np->min_hopcount = val;
 797                 break;
 798         case IPV6_DONTFRAG:
 799                 np->dontfrag = valbool;
 800                 retv = 0;
 801                 break;
 802         }
 803
 804         release_sock(sk);
 805
 806         return retv;
 807
 808 e_inval:
 809         release_sock(sk);
 810         return -EINVAL;
 811 }
 812
 813 int ipv6_setsockopt(struct sock *sk, int level, int optname,
 814                     char __user *optval, unsigned int optlen)
 815 {
 816         int err;
 817
 818         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 819                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 820
 821         if (level != SOL_IPV6)
 822                 return -ENOPROTOOPT;
 823
 824         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 825 #ifdef CONFIG_NETFILTER
 826         /* we need to exclude all possible ENOPROTOOPTs except default case */
 827         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 828                         optname != IPV6_XFRM_POLICY) {
 829                 lock_sock(sk);
 830                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
 831                                 optlen);
 832                 release_sock(sk);
 833         }
 834 #endif
 835         return err;
 836 }
 837
 838 EXPORT_SYMBOL(ipv6_setsockopt);
 839
 840 #ifdef CONFIG_COMPAT
 841 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
 842                            char __user *optval, unsigned int optlen)
 843 {
 844         int err;
 845
 846         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
 847                 if (udp_prot.compat_setsockopt != NULL)
 848                         return udp_prot.compat_setsockopt(sk, level, optname,
 849                                                           optval, optlen);
 850                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 851         }
 852
 853         if (level != SOL_IPV6)
 854                 return -ENOPROTOOPT;
 855
 856         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
 857                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
 858                         ipv6_setsockopt);
 859
 860         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 861 #ifdef CONFIG_NETFILTER
 862         /* we need to exclude all possible ENOPROTOOPTs except default case */
 863         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 864             optname != IPV6_XFRM_POLICY) {
 865                 lock_sock(sk);
 866                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
 867                                            optval, optlen);
 868                 release_sock(sk);
 869         }
 870 #endif
 871         return err;
 872 }
 873
 874 EXPORT_SYMBOL(compat_ipv6_setsockopt);
 875 #endif
 876
 877 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
 878                                   int optname, char __user *optval, int len)
 879 {
 880         struct ipv6_opt_hdr *hdr;
 881
 882         if (!opt)
 883                 return 0;
 884
 885         switch(optname) {
 886         case IPV6_HOPOPTS:
 887                 hdr = opt->hopopt;
 888                 break;
 889         case IPV6_RTHDRDSTOPTS:
 890                 hdr = opt->dst0opt;
 891                 break;
 892         case IPV6_RTHDR:
 893                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
 894                 break;
 895         case IPV6_DSTOPTS:
 896                 hdr = opt->dst1opt;
 897                 break;
 898         default:
 899                 return -EINVAL; /* should not happen */
 900         }
 901
 902         if (!hdr)
 903                 return 0;
 904
 905         len = min_t(unsigned int, len, ipv6_optlen(hdr));
 906         if (copy_to_user(optval, hdr, len))
 907                 return -EFAULT;
 908         return len;
 909 }
 910
 911 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 912                     char __user *optval, int __user *optlen)
 913 {
 914         struct ipv6_pinfo *np = inet6_sk(sk);
 915         int len;
 916         int val;
 917
 918         if (ip6_mroute_opt(optname))
 919                 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
 920
 921         if (get_user(len, optlen))
 922                 return -EFAULT;
 923         switch (optname) {
 924         case IPV6_ADDRFORM:
 925                 if (sk->sk_protocol != IPPROTO_UDP &&
 926                     sk->sk_protocol != IPPROTO_UDPLITE &&
 927                     sk->sk_protocol != IPPROTO_TCP)
 928                         return -ENOPROTOOPT;
 929                 if (sk->sk_state != TCP_ESTABLISHED)
 930                         return -ENOTCONN;
 931                 val = sk->sk_family;
 932                 break;
 933         case MCAST_MSFILTER:
 934         {
 935                 struct group_filter gsf;
 936                 int err;
 937
 938                 if (len < GROUP_FILTER_SIZE(0))
 939                         return -EINVAL;
 940                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
 941                         return -EFAULT;
 942                 if (gsf.gf_group.ss_family != AF_INET6)
 943                         return -EADDRNOTAVAIL;
 944                 lock_sock(sk);
 945                 err = ip6_mc_msfget(sk, &gsf,
 946                         (struct group_filter __user *)optval, optlen);
 947                 release_sock(sk);
 948                 return err;
 949         }
 950
 951         case IPV6_2292PKTOPTIONS:
 952         {
 953                 struct msghdr msg;
 954                 struct sk_buff *skb;
 955
 956                 if (sk->sk_type != SOCK_STREAM)
 957                         return -ENOPROTOOPT;
 958
 959                 msg.msg_control = optval;
 960                 msg.msg_controllen = len;
 961                 msg.msg_flags = 0;
 962
 963                 lock_sock(sk);
 964                 skb = np->pktoptions;
 965                 if (skb)
 966                         atomic_inc(&skb->users);
 967                 release_sock(sk);
 968
 969                 if (skb) {
 970                         int err = datagram_recv_ctl(sk, &msg, skb);
 971                         kfree_skb(skb);
 972                         if (err)
 973                                 return err;
 974                 } else {
 975                         if (np->rxopt.bits.rxinfo) {
 976                                 struct in6_pktinfo src_info;
 977                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
 978                                         np->sticky_pktinfo.ipi6_ifindex;
 979                                 np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
 980                                         ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
 981                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 982                         }
 983                         if (np->rxopt.bits.rxhlim) {
 984                                 int hlim = np->mcast_hops;
 985                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 986                         }
 987                         if (np->rxopt.bits.rxoinfo) {
 988                                 struct in6_pktinfo src_info;
 989                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
 990                                         np->sticky_pktinfo.ipi6_ifindex;
 991                                 np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
 992                                         ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
 993                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 994                         }
 995                         if (np->rxopt.bits.rxohlim) {
 996                                 int hlim = np->mcast_hops;
 997                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 998                         }
 999                 }
1000                 len -= msg.msg_controllen;
1001                 return put_user(len, optlen);
1002         }
1003         case IPV6_MTU:
1004         {
1005                 struct dst_entry *dst;
1006
1007                 val = 0;
1008                 rcu_read_lock();
1009                 dst = __sk_dst_get(sk);
1010                 if (dst)
1011                         val = dst_mtu(dst);
1012                 rcu_read_unlock();
1013                 if (!val)
1014                         return -ENOTCONN;
1015                 break;
1016         }
1017
1018         case IPV6_V6ONLY:
1019                 val = np->ipv6only;
1020                 break;
1021
1022         case IPV6_RECVPKTINFO:
1023                 val = np->rxopt.bits.rxinfo;
1024                 break;
1025
1026         case IPV6_2292PKTINFO:
1027                 val = np->rxopt.bits.rxoinfo;
1028                 break;
1029
1030         case IPV6_RECVHOPLIMIT:
1031                 val = np->rxopt.bits.rxhlim;
1032                 break;
1033
1034         case IPV6_2292HOPLIMIT:
1035                 val = np->rxopt.bits.rxohlim;
1036                 break;
1037
1038         case IPV6_RECVRTHDR:
1039                 val = np->rxopt.bits.srcrt;
1040                 break;
1041
1042         case IPV6_2292RTHDR:
1043                 val = np->rxopt.bits.osrcrt;
1044                 break;
1045
1046         case IPV6_HOPOPTS:
1047         case IPV6_RTHDRDSTOPTS:
1048         case IPV6_RTHDR:
1049         case IPV6_DSTOPTS:
1050         {
1051
1052                 lock_sock(sk);
1053                 len = ipv6_getsockopt_sticky(sk, np->opt,
1054                                              optname, optval, len);
1055                 release_sock(sk);
1056                 /* check if ipv6_getsockopt_sticky() returns err code */
1057                 if (len < 0)
1058                         return len;
1059                 return put_user(len, optlen);
1060         }
1061
1062         case IPV6_RECVHOPOPTS:
1063                 val = np->rxopt.bits.hopopts;
1064                 break;
1065
1066         case IPV6_2292HOPOPTS:
1067                 val = np->rxopt.bits.ohopopts;
1068                 break;
1069
1070         case IPV6_RECVDSTOPTS:
1071                 val = np->rxopt.bits.dstopts;
1072                 break;
1073
1074         case IPV6_2292DSTOPTS:
1075                 val = np->rxopt.bits.odstopts;
1076                 break;
1077
1078         case IPV6_TCLASS:
1079                 val = np->tclass;
1080                 break;
1081
1082         case IPV6_RECVTCLASS:
1083                 val = np->rxopt.bits.rxtclass;
1084                 break;
1085
1086         case IPV6_FLOWINFO:
1087                 val = np->rxopt.bits.rxflow;
1088                 break;
1089
1090         case IPV6_RECVPATHMTU:
1091                 val = np->rxopt.bits.rxpmtu;
1092                 break;
1093
1094         case IPV6_PATHMTU:
1095         {
1096                 struct dst_entry *dst;
1097                 struct ip6_mtuinfo mtuinfo;
1098
1099                 if (len < sizeof(mtuinfo))
1100                         return -EINVAL;
1101
1102                 len = sizeof(mtuinfo);
1103                 memset(&mtuinfo, 0, sizeof(mtuinfo));
1104
1105                 rcu_read_lock();
1106                 dst = __sk_dst_get(sk);
1107                 if (dst)
1108                         mtuinfo.ip6m_mtu = dst_mtu(dst);
1109                 rcu_read_unlock();
1110                 if (!mtuinfo.ip6m_mtu)
1111                         return -ENOTCONN;
1112
1113                 if (put_user(len, optlen))
1114                         return -EFAULT;
1115                 if (copy_to_user(optval, &mtuinfo, len))
1116                         return -EFAULT;
1117
1118                 return 0;
1119                 break;
1120         }
1121
1122         case IPV6_TRANSPARENT:
1123                 val = inet_sk(sk)->transparent;
1124                 break;
1125
1126         case IPV6_RECVORIGDSTADDR:
1127                 val = np->rxopt.bits.rxorigdstaddr;
1128                 break;
1129
1130         case IPV6_UNICAST_HOPS:
1131         case IPV6_MULTICAST_HOPS:
1132         {
1133                 struct dst_entry *dst;
1134
1135                 if (optname == IPV6_UNICAST_HOPS)
1136                         val = np->hop_limit;
1137                 else
1138                         val = np->mcast_hops;
1139
1140                 if (val < 0) {
1141                         rcu_read_lock();
1142                         dst = __sk_dst_get(sk);
1143                         if (dst)
1144                                 val = ip6_dst_hoplimit(dst);
1145                         rcu_read_unlock();
1146                 }
1147
1148                 if (val < 0)
1149                         val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1150                 break;
1151         }
1152
1153         case IPV6_MULTICAST_LOOP:
1154                 val = np->mc_loop;
1155                 break;
1156
1157         case IPV6_MULTICAST_IF:
1158                 val = np->mcast_oif;
1159                 break;
1160
1161         case IPV6_MTU_DISCOVER:
1162                 val = np->pmtudisc;
1163                 break;
1164
1165         case IPV6_RECVERR:
1166                 val = np->recverr;
1167                 break;
1168
1169         case IPV6_FLOWINFO_SEND:
1170                 val = np->sndflow;
1171                 break;
1172
1173         case IPV6_ADDR_PREFERENCES:
1174                 val = 0;
1175
1176                 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1177                         val |= IPV6_PREFER_SRC_TMP;
1178                 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1179                         val |= IPV6_PREFER_SRC_PUBLIC;
1180                 else {
1181                         /* XXX: should we return system default? */
1182                         val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1183                 }
1184
1185                 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1186                         val |= IPV6_PREFER_SRC_COA;
1187                 else
1188                         val |= IPV6_PREFER_SRC_HOME;
1189                 break;
1190
1191         case IPV6_MINHOPCOUNT:
1192                 val = np->min_hopcount;
1193                 break;
1194
1195         case IPV6_DONTFRAG:
1196                 val = np->dontfrag;
1197                 break;
1198
1199         default:
1200                 return -ENOPROTOOPT;
1201         }
1202         len = min_t(unsigned int, sizeof(int), len);
1203         if(put_user(len, optlen))
1204                 return -EFAULT;
1205         if(copy_to_user(optval,&val,len))
1206                 return -EFAULT;
1207         return 0;
1208 }
1209
1210 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1211                     char __user *optval, int __user *optlen)
1212 {
1213         int err;
1214
1215         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1216                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1217
1218         if(level != SOL_IPV6)
1219                 return -ENOPROTOOPT;
1220
1221         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1222 #ifdef CONFIG_NETFILTER
1223         /* we need to exclude all possible ENOPROTOOPTs except default case */
1224         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1225                 int len;
1226
1227                 if (get_user(len, optlen))
1228                         return -EFAULT;
1229
1230                 lock_sock(sk);
1231                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1232                                 &len);
1233                 release_sock(sk);
1234                 if (err >= 0)
1235                         err = put_user(len, optlen);
1236         }
1237 #endif
1238         return err;
1239 }
1240
1241 EXPORT_SYMBOL(ipv6_getsockopt);
1242
1243 #ifdef CONFIG_COMPAT
1244 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1245                            char __user *optval, int __user *optlen)
1246 {
1247         int err;
1248
1249         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1250                 if (udp_prot.compat_getsockopt != NULL)
1251                         return udp_prot.compat_getsockopt(sk, level, optname,
1252                                                           optval, optlen);
1253                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1254         }
1255
1256         if (level != SOL_IPV6)
1257                 return -ENOPROTOOPT;
1258
1259         if (optname == MCAST_MSFILTER)
1260                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1261                         ipv6_getsockopt);
1262
1263         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1264 #ifdef CONFIG_NETFILTER
1265         /* we need to exclude all possible ENOPROTOOPTs except default case */
1266         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1267                 int len;
1268
1269                 if (get_user(len, optlen))
1270                         return -EFAULT;
1271
1272                 lock_sock(sk);
1273                 err = compat_nf_getsockopt(sk, PF_INET6,
1274                                            optname, optval, &len);
1275                 release_sock(sk);
1276                 if (err >= 0)
1277                         err = put_user(len, optlen);
1278         }
1279 #endif
1280         return err;
1281 }
1282
1283 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1284 #endif
1285